Two very different types of models of visual perception have been used by researchers in the fields of cognitive psychology and human information processing. Discrete models characterized visual perception as a discrete operation that must completely finish processing a stimulus before any further cognitive processes receive information about that stimulus (e.g., Sternberg, 1969a). Continuous models characterize it as a continuous operation that gradually transmits information about a stimulus to later cognitive processes, so that these processes can begin before perception is finished (e.g., McClelland, 1979). It is important for both theoretical and methodological reasons to be able to discriminate between these two types of models, and the goal of the proposed research is to develop experimental tests that can choose between them. The proposed experiments emphasize two lines of research. One set of experiments will use psychophysiological measurements to look for evidence that some response preparation has occurred before stimulus perception has finished. Previous studies have sought evidence of such preliminary response preparation using indirect measures based on total response latency, and results have been subject to alternative explanations. Using electrophysiological measures including the electromyogram (EMG) and event-related cerebral potentials (ERPs), it should be possible to measure preliminary response preparation more directly. These measures will be obtained in several different experimental paradigms yielding effects previously attributed to preliminary response preparation, and they will indicate whether such preparation is actually responsible for these effects. Such preliminary response preparation would be consistent with continuous, but not discrete, models of perceptual processes. The second set of experiments will use reaction time measurements to study visual attention. If it can be established that attentional emphasis continuously modulates the output of perceptual processes (rather than their use by late processes), then this would constitute broad support for models with continuous rather than discrete activation of stimulus codes. The results of these studies will be of interest to cognitive psychologists, human factors engineers and designers of interfaces for human/computer interaction, and researchers in some areas of artificial intelligence.